Forest condition has been monitored by crown condition assessments, in Norway and in most of. Europe since the mid 1980s. This activity is based on a concern ...
04 Solberg (jl/k)
8/9/99 10:46 am
Page 329
Crown condition of Norway spruce: within-stand relationships and competition S. SOLBERG AND E. MOSHAUG 1
Norwegian Forest Research Institute, Division of Forest Ecology, Høgskoleveien 12, N-1432 Ås, Norway
Summary This study describes how crown density changes were distributed within monitoring plots, in order to determine whether the reduced crown density observed could be explained as a worsening of a limited number of unhealthy or small and slightly suppressed trees. Crown density, yellowing, coning and stem diameter data were available from 447 selected plots comprising 22 560 single trees all having a complete 1990–97 series of crown condition data. The 8-year record of crown density for each tree was recalculated to two median values, for 1990–93 and 1994–97, in order to reduce the influence from short-term variations including random errors. The scores for yellowing and number of cones were averaged over the years 1990–93. These variables, and diameter, were recalculated to rank indices within each plot. Relationships between variables were described by graphs and examined by correlation analyses of the indices. The trees tended to retain their internal ranking. Generally, when crown density for a plot has changed, most of the trees were affected. The most defoliated trees in each plot had the least negative changes, but apart from that the trees were equally affected regardless of their yellowing, number of cones, and their size. The results demonstrate that any effects from competition between the trees were sufficiently removed in the assessments even in densely stocked stands.
Introduction Forest condition has been monitored by crown condition assessments, in Norway and in most of Europe since the mid 1980s. This activity is based on a concern for negative effects of longrange air pollutants. The ‘Local county monitoring plots’ are a nationwide set of monitoring plots in Norway included in the Norwegian monitoring programme for forest damage (Horntvedt et al., 1992). During a decade of monitoring, a reduction of crown density has © Institute of Chartered Foresters, 1999
been noted in these plots (Groeggen, 1997). The reduction is relatively moderate, and the causes are unclear. Thus, there has been a need to determine if the reduction could be explained by natural processes in stand development. Based on these data, Solberg (1999) concluded that ageing is not a sufficient explanation for the reduction of crown density observed. A second aspect of stand development regards a possible influence of a few diseased or small and slightly suppressed trees, which may accumulate in the stand, as thinning operations are rare in today’s Forestry, Vol. 72, No. 4, 1999
04 Solberg (jl/k)
330
8/9/99 10:46 am
Page 330
F O R E S T RY
forestry. A plot mean value of crown density is sensitive to the influence of a few trees with low crown density, because all trees are given equal weight in calculations and a majority of the trees have fairly high scores. This can be put in a wider context dealing with how to select trees for measurements, in order to describe the development of a forest stand. If all trees in a stand are measured, those trees suffering from competition may severely influence the plot value and make it less representative for the stand development. On the other hand, if only the largest or most healthy trees are selected for measurements, another problem arises because of fluctuations in their performance (Cherubini et al., 1998). It is likely that the previous development of such trees is above the plot average, which is partly the reason why they are among the larger and healthier trees now, and later it is likely that their performance will gradually decrease towards the plot mean. Another influence on how crown density is distributed within plots is stress, and several stress factors could have caused the reduced crown density. The importance of a stress factor can be addressed by correlative studies using plot values, where the actual stress factor is quantified for each plot. An alternative approach is to study how the different trees in each plot are affected, by correlating crown condition, and other variables describing various tree features. For example, LeBlanc (1990) found that a growth decline of red spruce (Picea rubra Sarg.) in the Appalachians, USA, was most pronounced for larger, dominant trees, and for trees surrounded by few competitors, indicating a type of stress affecting exposed trees, rather than competition as the causal agent. The aim of this study was to determine whether the reduced crown density could be explained as a worsening of a limited number of unhealthy or small and slightly suppressed trees. The following questions were raised: Is reduced crown density affecting only a limited number of the trees within each plot?; Do the trees retain their internal ranking of crown density?; Is the reduction most severe for low crown density trees, yellowish trees, trees with heavy cone production, or the smaller trees?; Can competition between trees be of importance for reduced crown density?
Materials The local county monitoring plots were established throughout Norway mainly in 1988. In each forest officer’s district a cluster of four plots in different age and development stages were established: ‘III’ – young; ‘IV’ – intermediate; ‘V’ – old; and ‘Extreme’ – declining. The last of these was defined as plots having defoliated or yellowish crowns, where these symptoms could not readily be explained by growing conditions. Each plot should contain 50 non-suppressed trees after the final thinning, which gave variable plot sizes of about 0.05–0.1 ha. A number of selection criteria were defined before the establishment. However, as not all criteria could be met for all plots, they served as guidelines: the plots should preferably be located in productive stands; typical stands for the region; having Norway spruce (Picea abies (L.) Karst) as the major species; and the most common vegetation type, which is dominated by bilberry (Vaccinium myrtillus L.; Kielland-Lund, 1981); and with intermediate site index. Silvicultural activities have been limited to regular thinnings in a few type III plots. Local forest officers have selected and maintained the plots, and assessed crown condition on all trees annually in September. The assessments were made from the ground with binoculars, viewing the trees from different directions whenever possible, and all were subjective scorings. Such data are crude and error prone, however, in a control survey where 12 000 Norway spruce trees on these plots were independently reassessed it was found that 45 per cent of the trees had less than 5 per cent difference in crown density (Solberg and Strand, 1999). Crown density was defined as the mass of live foliage as a percentage of what is expected for the actual tree at that site, within the upper half of the live crown. Effects from competition should in principle be removed, by excluding parts of the crown that are directly influenced by neighbouring trees, and by taking the social position of the tree into account. Crown density corresponds to the term defoliation, which is widely used as the main crown feature in the European-wide forest monitoring (Lorenz, 1995). Crown colour refers to the amount of yellowish foliage, given in the classes: 1 – normal green (0–10 per cent yellow foliage); 2 – slight yellow (11–25 per cent); 3 – moderate yellow (26–60 per cent); and 4 – strong
04 Solberg (jl/k)
8/9/99 10:46 am
Page 331
C R O W N C O N D I T I O N O F N O RWAY S P R U C E
yellow (> 60 per cent). The number of cones is given in the classes: 1 – none or few; 2 – intermediate; and 3 – many. These classes have not been objectively defined, because coning has been of minor importance in the monitoring, and because coning in spruce trees often can be described as a Boolean variable. A control survey on the plots, where the cones were counted, showed that the observers have allocated roughly 0–50, 50–300 and > 300 cones to the classes respectively, with some adjustments for tree size. In 1996 stem diameter at breast height was measured for all trees having diameters above 5 cm. Heights were measured on a selection of trees, by which top height could be calculated as the mean height of the five trees with the largest stem diameters. The Norwegian monitoring programme for forest damage at Norwegian Forest Research Institute (NISK), has arranged field assessment training sessions, control surveys, data handling and reporting (e.g. Groeggen, 1997). For this study 447 plots comprising 22 560 single trees having a complete 8-year record of crown condition data from 1990 to 1997 were selected as follows: first, only plots where Norway spruce was the main species were included, and only trees of this species. Second, trees were excluded if they were affected by suppression, mechanical damage from snow or wind, or if they died or were cut during the study period. Third, plots were excluded if more than 15 per cent of the spruce trees lacked a complete series of assessments from unknown reasons. On average, the plots comprised 70 trees, of which 51 trees were included as sample trees, while the number of excluded trees were as follows: 1.3 were Scots pine (Pinus sylvestris L.); 1.9 were deciduous trees; 6.7 were suppressed; 6.6 were mechanically damaged, cut or dead; and 3.1 lacked observations of unknown reasons. The development of crown condition is given in Table 1. Crown density has been steadily reduced, while yellowing and the number of cones have shown year-to-year variations and no trend. This overall description of crown condition development, applies fairly well for most plots and regions in Norway, except western and northern Norway which have had clearly higher crown density, less negative development, and fewer yellowish trees and cones.
331
Table 1: Development of mean crown density, mean fraction of yellowish trees, and mean fraction of trees with cones
Year
Crown density
Crown colour
Cones
1990 1991 1992 1993 1994 1995 1996 1997
87.6 87.9 87.5 87.0 86.7 86.1 84.9 84.4
12 14 16 13 16 11 14 17
4 2 17 16 4 30 2 2
Methods The eight years under study were divided into two periods; 1990–93 and 1994–97. Each of the crown condition variables was recalculated to one value for the first period, as follows: • Crown density = median (crown densities from 1990 to 1993) • Yellowing = average (crown colour scores from 1990 to 1993) • Coning = average (cone scores from 1990 to 1993) These values should represent the initial health state of each tree, and the influence from shortterm variation and random errors should be minimized. Now, crown colour and number of cones are nominal variables, which in principle cannot be averaged. However, no other appropriate way to obtain a 1990–93 score for these variables was available. For crown density similar median values were calculated for the second period, and the crown density change was set to the difference between the two medians: Crown density change = crown density1994–972 crown density1990–93 The above variables, as well as crown density in 1990 and 1997, and stem diameter, were recalculated to rank indices between 0 and 1000, such as: Crown density rank index = 1000 * crown density rank order/number of trees on the plot By this, for each plot and each variable rank indices were produced, which varied from zero to
04 Solberg (jl/k)
332
8/9/99 10:46 am
Page 332
F O R E S T RY
1000 regardless of the variance of that variable. Some plots and some variables had very low variance, and with these the rank indices had to be mostly randomly assigned to the trees. This was especially the case for yellowing and coning. In order to describe the stability of the trees’ crown density ranking within the plots, Spearman rank order correlations were calculated between crown density ranks of 1990 and those of the following years. The development of crown density was described by graphs, in which the mean crown density of ten groups was given. Each group comprised 10 per cent of the trees from each plot, and in the first group were the trees which in 1990 had crown density ranks from zero to 100, in the next group those trees with ranks from 100 to 200 and so on. A similar graph was made after the trees were assigned to groups according to their crown density ranks in 1997. The rest of the results are based on the crown density medians. Spearman rank order correlations were calculated between the crown density medians from the first period and those of the second period. In order to search for relationships between the variables the rank indices were correlated by Spearman rank order correlations. This approach could only identify possible relationships, but provides little information about their magnitude. Thus, the relationships were visualized by graphs. These graphs were similar to the first graphs, however, they were designed in order to demonstrate the main results only. In each graph, crown density and its changes were presented as averages of the crown density medians in each period for the top 10 per cent of the trees (rank indices above 900) and for the bottom 10 per cent of the trees (rank indices below 100) as ordered by the actual variable. The averages for all trees were given as well. In order to give equal influence for all plots, regardless of the number of trees, the averages were first calculated for each plot, and then these averages were averaged again. Crown density changes also varied between plots, and the plots were divided in four groups. Each group comprised a quarter of the plots according to their crown density changes; the first group contained the quarter of plots with the most negative development and so on. The distribution of crown density changes for single trees were calculated for each of these groups.
If effects of competition among the trees was not sufficiently removed in the field assessments, and it is important for crown density development, the smaller trees should have lower crown densities, and the smaller and more defoliated trees should have more negative crown density changes. This should be most pronounced in densely stocked stands. Stand density increases both with the number of trees per area and their sizes (Hyink and Zedaker, 1987), and the HartBecking’s spacing index (Clutter et al., 1983) was used as a measure of stand density: Hart-Becking’s spacing index = average distance between the trees –––––––––––––––––––––––––––––– top height This index gives low values for dense stands, and self-thinning is expected at values below 0.10–0.12. The development of crown density and its relationships to stem diameter in dense stands (spacing index 0.12) were visualized by graphs. Effects of competition are often addressed using competition models that take into account the actual neighbours for each tree (Hyink and Zedaker, 1987), however those methods require co-ordinates for the location of the trees, which were lacking in this case.
Results Considerable variation was found between the plots, regarding the variables under study (Table 2). There were plots without defoliation at all, plots with hardly any yellowing or any cone production. There were plots with strong increases and plots with strong decreases of crown density. Mean stem diameters also varied considerably. This has two implications for the study: first, it is likely that plots exist which deviate considerably Table 2: Descriptive statistics for plot means
Crown density1990–93 Crown density1994–97 Crown density change Yellowing Coning Stem diameter
Average
Min–max
88 86 22.0 – 1.2 – 1.1 233
(47–99) (41–99) (218–14) (1.0–2.4) (1.0–1.6) (105–435)
04 Solberg (jl/k)
8/9/99 10:46 am
Page 333
C R O W N C O N D I T I O N O F N O RWAY S P R U C E
333
Table 3: Between- and within-plot variance: analysis of variance for the variables with factor plot
Crown density1990–93 Crown density1994–97 Crown density change Yellowing Coning Stem diameter
Between plot variance
Within plot variance
3502 4647 883 1.91 0.65 16 7062
50 64 26 0.054 0.018 2891
Crown density
from any general description of the results. Second, a number of plots have low, or zero, variance, and in the results these plots help to conceal within-stand relationships that are evident on other plots. All the variables under study had clear differences between plots, and the variance between plots was much higher than within plots (Table 3). The variance in crown density within plots increased from the first to the second period. The trees tended to retain their internal ranking and to develop parallel to each other (Figure 1). The large influence from short-term variation including random errors in crown density is evident: when the trees were ranked according to crown density in 1990, the differences between the rank groups were severely reduced to 1991. Similar results were obtained when trees were
10 9 8 90 76 5 4 3
90
80 2
80
10 9 8 7 6 5 4 3 2
1
70
70 1
90
92
94
96
90
92
94
96
Figure 1. Development of crown density 1990–97. The trees are assigned to ten groups according to their rank order of crown density in 1990 (left) and in 1997 within the plot. Each group consists of 10 per cent of the trees (about five trees) from each plot. Each group is given as a line representing the mean values of that group for each year.
F(446, 22 113) 70 72 34 35 36 58
ranked according to 1997 crown density. Rank order correlations between crown density in 1990 and crown density in 1991 was 0.50 (P < 0.05), and between 1990 and 1997 it was 0.35 (P < 0.05). The rest of the results are based on the crown density medians. A similar correlation was found between the crown density medians in the first period and those of the second period (r = 0.58, P < 0.05). When correlations were calculated for each plot, the coefficients ranged from zero to one. When crown density for a plot has changed, most of, or all trees were affected (Table 4). For the quarter of plots having the most negative development of crown density, each of the 4676 trees on those plots had a reduction below 22.5. In the next quarter of plots, all the trees had changes around 25 and zero. In the third group none of the trees had major changes. In the fourth group all trees had either minor changes or increases in crown density. The most defoliated trees in each plot had the least negative changes (Figure 2). A negative correlation was found between crown density and crown density changes (r = 20.24, P < 0.05). This was most pronounced for plots with positive changes. The quarter of plots with the most positive crown density changes had an average crown density change of +3, while the 10 per cent most defoliated trees on those plots had a change of +8. On the quarter of plots with the most negative changes there was an average change of 27 for the trees regardless of their crown density rank. Yellowing was associated with low crown density (Figure 2), and a negative correlation was found (r = 20.21, P < 0.05), but no correlation was found between yellowing and crown density changes. Coning was weakly correlated to crown density (r = 0.04, P < 0.05), but not to crown
04 Solberg (jl/k)
8/9/99 10:46 am
Page 334
334
F O R E S T RY
Table 4: Frequency distribution (%) of 5% classes of crown density changes, for four groups of plots grouped according to their average crown density changes. The classes are indicated by their midpoint values, and at both ends of the scale several classes are grouped Single tree changes ———————————————————————————————————— –70 to –25 –20 –15 –10 –5 0 5 10 15 20 25 to 45
Plot group
Plot change
1 2 3 4
–18 to –4.2 –4.2 to –1.7 –1.7 to 0.0 0.0 to 14
(a)
3.7
dense
4.0
9.8
26.4 56.1 18.5
(b)
81.5 100.5 59.8 28.2
8.3
(c)
2.5
0.7
0.5
(d)
10 green many cones 90
10 All
Crown density
All
1
high diameter 10 All 1
10 All 1
low diameter
few cones
80 yellow 1
70
defoliated
90–93
94–97
90–93
94–97
90–93
94–97
90–93
94–97
Figure 2. Development of crown density given as mean values of the medians in the first and the second period for the top and bottom 10 per cent of the trees (about five trees) at each plot ranked according to (a) crown density, (b) yellowing, (c) coning, and (d) stem diameter. Mean values for all trees are given as dotted lines.
density changes. Diameter was correlated to crown density (r = 0.11, P < 0.05), but not to crown density changes. The Hart-Becking’s spacing index varied from 0.09 to 0.39, with an average of 0.17. Densely stocked stands, having index values below 0.12 numbered 35. In these plots crown density changes were minor. Trees with high crown density had a decrease, and the most defoliated had an increase. Tree size (diameter) had a slightly higher influence on these plots. On average, the 10 per cent largest trees had a crown density of 90.5 in the first period and 91.3 in the second.
The corresponding figures for the 10 per cent smallest trees were 87.5 and 87.3.
Discussion Within the plots, the trees tended to retain their internal crown density ranking, and to develop parallel to each other. Considerable variations in the ranking of the trees from year to year are found in other studies (Hägi, 1989; Innes, 1993), but in the present study this was removed by using 4-year median values for each tree. There
04 Solberg (jl/k)
8/9/99 10:46 am
Page 335
C R O W N C O N D I T I O N O F N O RWAY S P R U C E
was also evidence from Figure 1 that changes in the rank orders took place. The correlation coefficient between crown density in 2 years was not too high, and it decreased with increasing number of years. Trees with yellowish crowns had a stable, lower crown density, but not more negative development of crown density. An explanation for these results is that both chronic diseases and acute stresses have been present, and that trees suffering from chronic diseases remain at a low and stable level of crown density, while trees affected by acute stress recover. Among chronic diseases, root rot may be important, which at advanced stages may lead to crown thinning and yellowing. Observer annotations and control surveys have revealed also some acute causes for yellowing and defoliation, including the rust fungus Chrysomyxa abietis (Unger) and heavy, autumnal senescent yellowing triggered by drought (Solberg et al., 1994). In these cases, yellowing is inevitably followed by needle loss. Long-term stability of trees’ health and growth ranking is found in other studies. The ranking of trees with regard to their level of defoliation or yellowing is found to correspond to stable differences in growth for several years or even up to a century (Röhle, 1987; Bert, 1993), but occasionally, the ranking with regard to growth and size may change considerably (Cherubini et al., 1998), as a result of severe stress such as drought (Innes and Neumann, 1991; Landmann, 1993). These studies demonstrated that the allocation of the trees to present levels of health scores occurred in certain years of severe stress. According to this, it is possible that the reduced crown density observed on the local county monitoring plots was determined by stress before the period under study, and that a number of trees now are predestined to a decline (Landmann, 1993). Even though all trees within a plot have experienced more or less the same development, the results clearly show that the plot development should be described by data for all trees. If a group of trees were selected for assessments, e.g. the apparently most healthy ones, their development would not be representative for the plot development (see Figure 1), which is in line with the results of Cherubini et al. (1998) concerning growth studies. The number of cones was poorly related to
335
crown density development, however, high numbers of cones were slightly associated with high crown density. Generally, cones are often referred to as a symptom of diseases like root rot. Possibly, unhealthy trees may carry a number of cones every year, and in the field this association will be evident in years when most trees have no cones. However, in the long run, the more vital trees may still produce more cones. Siegl and Shönborn (1990) and Innes (1994) also found a positive relationship between the number of cones and crown density. Competition between the trees, or crowding, is apparently of negligible importance for the low and decreasing crown density observed. Large and small trees have developed in a fairly similar way, even in densely stocked stands. This indicates that any effects of competition or crowding are sufficiently removed in the assessments. This important aspect of the field work is normally paid considerable attention during training sessions. The removal of data from suppressed trees contributes to the complete removal of competition effects. In contrast to this, Matejka (1995) found it necessary to adjust plot values of crown density in the Czech republic, in order to reduce the effects from a limited number of trees suffering from competition. In a highly polluted area in the same country, Vacek and Leps (1996) found the most negative development of crown density among suppressed trees, but argued that air pollution caused their decline. Reduced crown density could not be explained as a worsening of a limited number of unhealthy or small trees within each plot. Effects from competition were not found, not even in densely stocked stands. This result, and the finding that increased age was not a sufficient explanation (Solberg, 1999), demonstrate that the reduction in crown density is not caused by natural stand development. The major stress factors remain to be determined.
Acknowledgement Thanks to Kåre Venn for comments to the manuscript. The efforts of the local forest officers throughout Norway are acknowledged. The study was financed by the Ministry of Agriculture and the Ministry of Environment Protection as a part of the Norwegian Monitoring Programme for Forest Damage.
04 Solberg (jl/k)
8/9/99 10:46 am
336
Page 336
F O R E S T RY
References Bert, G.D. 1993 Impact of ecological factors, climatic stresses, and pollution on growth and health of silver fir (Abies alba Mill.) in the Jura mountains: an ecological and dendrochronological study. Acta Oecol. 14(2), 229–226. Cherubini, P., Dobbertin, M. and Innes, J.L. 1998 Potential sampling bias in long-term forest growth trends reconstructed from tree rings: A case study from the Italian alps. For. Ecol. Manage. 109(1–3), 103–119. Clutter, J.L., Fortson, J.C., Pienaar, L.V., Brister, G.H. and Balley, R.L. 1983 Timber Management: A Quantitative Approach. J. Wiley, New York, 333pp. Groeggen, T. 1997 Fylkesvise lokale overvåkingsflater. vitalitetsregistreringer 1997. [Local county monitoring plots. vitality survey 1997.] Rapport fra Skogforsk 8/97, 15pp. [In Norwegian, with English summary] Hägi, K. 1989 Terrestrische Waldschadeninventur. Bericht 314. Eidgenössische Anstalt für das forstliche Versuchswesen, Birmensdorf, 36pp. Horntvedt, R., Aamlid, D., Rørå, A. and Joranger, E. 1992 Monitoring programme for forest damage. An overview of the Norwegian programme. Norw. J. Agric. Sci. 6, 1–17. Hyink, D.M. and Zedaker, S.M. 1987 Stand dynamics and the evaluation of forest decline. Tree Physiol. 3(1), 17–26. Innes, J.L. 1993 Forest health: Its Assessment and Status. CAB International, Wallingford, 677pp. Innes, J.L. 1994 The occurrence of flowering and fruiting on individual trees over 3 years and their effects on subsequent crown condition. Trees Struct. Funct. 8(3), 139–150. Innes, J.L. and Neumann, H. 1991 Past growth variations in Picea sitchensis with differing crown densities. Scand. J. For. Res. 6(3), 395–406. Kielland-Lund, J. 1981 Die Waldgesellschaften SONorwegens. Phytocoenologia 9(1/2), 53–250. Landmann, G. 1993 Role of climate, stand dynamics and past management in forest declines: a review of ten years of field ecology in France. In: Hüttl/MüllerDombois (eds). Forest Decline in the Atlantic and Pacific Region. Springer-Verlag, Berlin.
LeBlanc, D.C. 1990 Red spruce decline on Whiteface Mountain, New York. I. Relationships with elevation, tree age, and competition. Can. J. For. Res. 20, 1408–1414. Lorenz, M. 1995 International co-operative programme on assessment and monitoring of air pollution effects on forests – ICP forests. Water Air Soil Pollut. 85(3), 1221–1226. Matejka, K. 1995 Monitoring of the forest status in the Czech Republic. Recent results and prospects. In Investigation of the Forest Ecosystems and of Forest Damage. Lowland and Submountain Forests and Monitoring of the Forest Status. K. Matejka (ed.). Proceedings of the workshop held in Kostelec nad Cernymi lesy on April 5–7, 1993, 146–154. Röhle, H. 1987 Entwiklung von Vitalität, Zuwach und Biomassenstruktur der Fichte in Verschiedenen bayerischen Untersuchungsgebieten unter dem Einfluss der neuartigen Walderkrankungen. Shriftenreihe der forstwissenschaftlichen Fakultät der Universität München und der bayer. forstlichen Versuchs- und Forschungsanstalt 83, 1–122. Siegl, T. and Schönborn, A.V. 1990 Fruktifikation der Fichte ( Picea abies (L.) Karsten). Forstwissenschaftlichen Fakultät der Ludvig-Maximillians-Universität München, 50pp. Solberg, S. 1990 Crown density changes of Norway spruce and the influence from increased age on permanent monitoring plots in Norway during 1988–97. Eur. J. For. Pathol. 29, 219–230. Solberg, S. and Strand, L. 1999 Crown density assessments, control surveys and reproducibility. Environ. Monit. Assess. 56, 75–86. Solberg, S., Solheim, H., Venn, K. and Aamlid, D. 1994 Tilfeller av skogskader i Norge i 1992 og 1993. [Cases of forest damage in Norway 1992 and 1993.] Rapport fra Skogforsk. 24/94, 1–35. [In Nor-wegian, with English summary] Vacek, S. and Leps, J. 1996 Spatial dynamics of forest decline: the role of neighbouring trees. J. Veg. Sci. 7, 789–798.
Received 16 July 1998
